CableDex: Cable Length Estimation on Industrial Reels Using a Handheld Device

Imagine you are standing in a giant warehouse filled with massive spools of cable, like the kind used to connect cities or power factories. These spools are heavy, and the cable is wound tightly around them. Now, picture a worker who needs to know exactly how much cable is left on a spool. In the old days, they had to guess based on how heavy the spool felt, or worse, they had to unroll the entire thing to measure it—a slow, messy, and frustrating process. This is the world of industrial manufacturing, where time is money and mistakes cost a lot.

To solve this, scientists use a branch of computer science called computer vision. Think of this as teaching a computer to "see" and understand the world through a camera lens, just like your eyes do. Two key tricks help the computer do this: segmentation, which is like a digital pair of scissors that can cut out a specific object (like a spool) from the background of a photo, and pose estimation, which is the computer's ability to figure out how an object is tilted or turned in 3D space just by looking at a flat picture. When you combine these tricks with a little bit of geometry, a computer can look at a single photo and calculate the volume of an object, even if it's a complex shape like a coil of wire. This paper tackles the messy, real-world problem of applying these high-tech tricks to a dusty factory floor to save workers time and headaches.

The Magic App That Counts Cables

Meet CableDex, a clever new system that acts like a digital detective for industrial cable reels. The problem it solves is simple but annoying: when cable is pulled off a giant reel, nobody knows exactly how much is left without doing a lot of hard work. CableDex changes the game by letting a worker take just one photo of a reel with a regular smartphone, and then instantly telling them a highly accurate estimate of the remaining cable length.

Here is how the magic happens, step-by-step:

1. The "Digital Scissors" (Instance Segmentation)
First, the system needs to know what it is looking at. It uses a super-smart AI model (a type of neural network called YOLO11s) that has been trained on 1,000 photos of different cable reels. Think of this model as a pair of incredibly precise digital scissors. When you snap a picture, the AI instantly cuts out the reel and the cable from the rest of the photo, creating a perfect outline. It's so good at this that it gets it right 99.5% of the time, and it does the whole job in just 5.66 milliseconds—faster than you can blink.

2. The "Smart Eye" (Camera Calibration)
Before the computer can measure anything, it needs to know how the phone's camera sees the world. Every phone lens is slightly different, like how every human eye sees colors a tiny bit differently. CableDex has a built-in "memory bank" of calibration data for different phone models. If the phone is new to the system, the user films a quick five-second video of a checkered pattern (like a chessboard). The system analyzes this video to figure out the camera's "inner secrets" (its focal length and lens distortion) so it can measure things accurately later.

3. The "3D Detective" (Pose Estimation)
Now that the computer knows what the reel is and how the camera sees it, it plays a game of 3D geometry. It looks at the four corners of the reel in the photo and compares them to the reel's known real-world size. By doing this, it can figure out exactly how the reel is sitting in space—how far away it is and how it is tilted. This is called pose estimation. Once it knows the reel's position, it can map the cable on top of it into 3D space, calculating the top radius, bottom radius, and height of the cable pile.

4. The "Volume Calculator" (The Math Part)
With the 3D shape of the cable pile figured out, the system uses a formula for a "truncated cone" (a cone with the top cut off) to calculate the total volume of the cable. But there's a catch: cables aren't solid blocks of plastic; they are round wires with tiny air gaps between them. The system accounts for this by subtracting about 21.5% of the volume to represent the air between the wires. Finally, it divides this volume by the thickness of the cable (which it gets from scanning the reel's barcode label) to reveal the total length.

Does It Actually Work?

The team tested CableDex in the real world at a factory in Portugal. They didn't just simulate it on a computer; they went on-site and tested it on 75 different reels across five different types of spools.

The results were impressive. The system estimated the cable length with an average error of just 4.90%. In the industrial world, an error margin of 10% is usually considered acceptable, so CableDex is comfortably within the safe zone. The team found that the size of the reel didn't matter; the system was just as accurate on small reels as it was on huge ones.

The Catch and the Future

While the system is a huge improvement over guessing by weight or unrolling cables, the authors are honest about its limits. The math assumes the cable is wound perfectly and neatly, like a stack of pancakes. If the cable is wound loosely or messily, the "truncated cone" shape changes, and the estimate might be less accurate.

However, for the vast majority of factory situations, CableDex offers a fast, reliable way to turn a simple photo into a precise measurement. It replaces slow, manual guesswork with a quick snapshot, proving that sometimes, the best way to measure a giant industrial problem is just to take a picture.